Denitrifying bacteria quantitative feeding method and system for aquaculture

Abstract

The invention discloses a denitrifying bacteria quantitative putting method and system for aquaculture, and belongs to the technical field of aquaculture, and the technical scheme is characterized by comprising the following steps: obtaining the excretion amount and the bait residual amount of each type of fingerling in each fish pond within a fixed time; according to the excretion amount, the bait remaining amount and the long-short time sequence model, the ammonia nitrogen content in each fish pond within the preset time is predicted; according to the ammonia nitrogen content, a bacterium throwing pond and the bacterium throwing amount and the bacterium throwing type of the bacterium throwing pond are determined in all the fish ponds; according to the bacterium throwing amount and the bacterium throwing type, quantitative bacterium throwing of the bacterium throwing pool is completed. In the aquaculture process, ammonia nitrogen and nitrite can affect respiration, immunity and growth of fishes, denitrifying bacteria can reduce the concentration of the substances and maintain balance of a water body ecological system, and therefore the ammonia nitrogen content in all the fish ponds is predicted through the feeding amount, and the water body ecological system is predicted through the feeding amount. The bacterium feeding amount and bacterium feeding types are reasonably planned according to the types and growth stages of the fingerlings in the fishponds, and a suitable environment is provided for fingerling breeding.

Description

Technical Field

[0001] The present invention relates to the technical field of aquaculture, and more specifically to a method and system for quantitatively dosing denitrifying bacteria for aquaculture. Background Art

[0002] In the field of aquaculture, the healthy reproduction of fish species is a key link to ensure aquaculture benefits and the sustainable development of the industry. As one of the core factors affecting fish species reproduction, water quality directly affects the growth, gonadal maturity, and reproduction success rate of fish species. As an important microorganism for regulating water quality, denitrifying bacteria can convert nitrogen-containing harmful substances such as ammonia nitrogen and nitrite, which are highly toxic to fish species in water, into nitrogen gas through nitrification and denitrification, effectively reducing water nitrogen pollution and creating a stable and suitable reproduction environment for fish species. However, at present, the dosing method of denitrifying bacteria in aquaculture is mostly an empirical and extensive dosing method, lacking precise quantitative control means, so there are deficiencies in the existing technology. Summary of the Invention

[0003] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide a method and system for quantitatively dosing denitrifying bacteria for aquaculture. By analyzing basic data such as the feeding amount and combining long and short time series models to determine the dosing amount of bacteria, the cost can be effectively reduced.

[0004] To achieve the above purpose, the present invention provides the following technical solutions:

[0005] The present invention provides a method for quantitatively dosing denitrifying bacteria for aquaculture, including:

[0006] Obtaining the excretion amount and remaining bait amount of each type of fish species in each fish pond within a fixed time;

[0007] Predicting the ammonia nitrogen content in each fish pond within a preset time according to the excretion amount, the remaining bait amount, and a long and short time series model;

[0008] Determining the dosing fish ponds, the dosing amount and dosing types of bacteria in the dosing fish ponds according to the ammonia nitrogen content;

[0009] Completing the quantitative dosing of bacteria in the dosing fish ponds according to the dosing amount and dosing types of bacteria.

[0010] As a further improvement of the present invention, the obtaining the excretion amount and remaining bait amount of each type of fish species in each fish pond within a fixed time includes:

[0011] Obtaining the feeding amount, feeding rate, absorption rate of each type of fish species in each fish pond within the fixed time, and the cleaning times and water change times of each fish pond;

[0012] Based on the feeding amount, the feeding rate, the cleaning frequency, and the water change frequency, the remaining bait amount is obtained;

[0013] Based on the feeding amount, the absorption rate, the cleaning frequency, and the water change frequency, the excretion amount is obtained.

[0014] As a further improvement of the present invention, based on the excretion amount, the remaining bait amount, and the long short-term memory model, the ammonia nitrogen content in each fish pond within a preset time is predicted, including:

[0015] Based on historical aquaculture data, the long short-term memory model corresponding to each type of fish species is trained;

[0016] The excretion amount and the remaining bait amount of each type of fish species in each fish pond within the fixed time are input into the long short-term memory model corresponding to each type of fish species, and the ammonia nitrogen content in each fish pond within the preset time is predicted.

[0017] As a further improvement of the present invention, the training of the long short-term memory model corresponding to each type of fish species based on historical aquaculture data includes:

[0018] The historical aquaculture data is divided into a validation set and a test set;

[0019] Based on the validation set, the initial long short-term memory model is trained to obtain an optimized long short-term memory model;

[0020] Based on the test set, the optimized long short-term memory model is evaluated to obtain the long short-term memory model corresponding to each type of fish species.

[0021] As a further improvement of the present invention, the training of the initial long short-term memory model based on the validation set to obtain an optimized long short-term memory model includes:

[0022] Based on historical aquaculture data, a relationship curve between the excretion amount, the remaining bait amount, and the ammonia nitrogen content is obtained;

[0023] Based on the relationship curve, a prediction training set is obtained;

[0024] Based on the prediction training set, the initial long short-term memory model is trained to obtain a training model;

[0025] Based on the training model and the validation set, the optimized long short-term memory model is obtained.

[0026] As a further improvement of the present invention, the obtaining of the optimized long short-term memory model based on the training model and the validation set includes:

[0027] Divide the parameters in the training model into multiple parameter groups, and arrange the multiple parameter groups in sequence to obtain a first queuing sequence;

[0028] Perform a first iterative operation, where the first iterative operation includes solidifying all parameter groups except the first parameter group, updating the parameters in the current training model according to the validation set until the training model meets the first termination condition, updating the first queuing sequence and repeating the above solidifying and updating steps until the first queuing sequence is an empty sequence, and outputting the current training model as the optimized long short-term sequence model; the first parameter group is the parameter group at the head of the first queuing sequence.

[0029] As a further improvement of the present invention, training the initial long short-term sequence model according to the prediction training set to obtain a training model, including:

[0030] Divide the prediction training set into multiple training groups, and arrange the multiple training groups in sequence to obtain a second queuing sequence;

[0031] Perform a second iterative operation, where the second iterative operation includes inputting the first training group into the current long short-term sequence model for forward propagation calculation, calculating a loss function according to the forward propagation calculation result and the historical aquaculture data, calculating the gradient of the parameters in the current long short-term sequence model according to the loss function and the backpropagation algorithm, updating the current long short-term sequence model according to the gradient to obtain an updated long short-term sequence model, updating the second queuing sequence until the second termination condition is met, and outputting the current updated long short-term sequence model as the training model; the first training group is the training group at the head of the second queuing sequence; the loss function is a regularization loss function.

[0032] As a further improvement of the present invention, determining the bacteria injection pool, the bacteria injection amount and the bacteria injection type in each fish pond according to the ammonia nitrogen content, including:

[0033] Determine the bacteria injection pool according to the ammonia nitrogen content and a preset threshold;

[0034] Determine the bacteria injection type according to the type and growth stage of the fish species in the bacteria injection pool, and the bacteria injection type includes nitrifying bacteria, denitrifying bacteria and anaerobic ammonium oxidation bacteria;

[0035] Calculate the bacteria injection amount according to the nitrogen removal ability, action time and safety factor corresponding to the bacteria injection type.

[0036] As a further improvement of the present invention, completing the quantitative bacteria injection into the bacteria injection pool according to the bacteria injection amount and the bacteria injection type, including:

[0037] Determine the comprehensive score of the bacteria injection pond according to the bacteria injection amount, bacteria injection type, fish species type and location of the bacteria injection pond;

[0038] Complete the quantitative bacteria injection of the bacteria injection pond in sequence according to the comprehensive score.

[0039] On the other hand, the present invention provides a method and system for quantitatively injecting denitrifying bacteria for aquaculture, including:

[0040] An acquisition module, configured to acquire the excretion amount and bait remaining amount of each type of fish species in each fish pond within a fixed time;

[0041] A calculation module, configured to predict the ammonia nitrogen content in each fish pond within a preset time according to the excretion amount, the bait remaining amount and the long short-term memory model, and determine the bacteria injection pond, the bacteria injection amount and the bacteria injection type of the bacteria injection pond in each fish pond according to the ammonia nitrogen content;

[0042] A control module, configured to complete the quantitative bacteria injection of the bacteria injection pond according to the bacteria injection amount and the bacteria injection type.

[0043] The present invention predicts the ammonia nitrogen content in each fish pond through the bait feeding amount, and reasonably plans the bacteria injection amount and the bacteria injection type in combination with the fish species type and the growth stage in each fish pond, which can provide a suitable environment for fish species reproduction. Moreover, compared with the method of determining the ammonia nitrogen content through water quality detection, the method provided by the present invention can effectively reduce the detection cost and improve the decision-making efficiency. Brief Description of the Drawings

[0044] Figure 1 It is a schematic diagram of the method steps of the present invention;

[0045] Figure 2 It is a schematic diagram of the steps for training the initial long short-term memory model of the present invention;

[0046] Figure 3 It is a schematic diagram of the implementation scenario of the present invention Figure 1 ;

[0047] Figure 4 It is a schematic diagram of the implementation scenario of the present invention Figure 2 ;

[0048] Figure 5 It is a schematic diagram of the implementation scenario of the present invention Figure 3 。 Detailed Embodiments

[0049] The technical solution of the present invention will be described in detail below through the drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention.

[0050] The term "and / or" in the following text is merely a description of the relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Additionally, the character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0051] As Figure 1 shown, an embodiment of the present application provides a method for quantitatively dosing denitrifying bacteria for aquaculture, including:

[0052] Obtaining the excretion amount and remaining bait amount of each type of fish species in each fish pond within a fixed time;

[0053] Predicting the ammonia nitrogen content in each fish pond within a preset time according to the excretion amount, remaining bait amount, and long and short time series model;

[0054] Determining the dosing ponds, the dosing amount and dosing types of bacteria in the dosing ponds according to the ammonia nitrogen content;

[0055] Completing the quantitative dosing of bacteria in the dosing ponds according to the dosing amount and dosing types of bacteria.

[0056] Among them, the fixed time and the preset time are determined according to actual needs. For example, the fixed time can be designed as 30 days, that is, obtaining the daily excretion amount and remaining bait amount of each type of fish species in each fish pond within 30 days, and the preset time can be designed as 3 days or 5 days.

[0057] This embodiment can predict the ammonia nitrogen content in each fish pond in real time through the feeding amount, and reasonably plan the dosing amount and dosing types of bacteria in combination with the fish species type and growth stage in each fish pond, which can provide a suitable environment for fish species reproduction. Moreover, compared with the method of determining the ammonia nitrogen content through water quality detection, the method provided by the present invention can effectively reduce the detection cost and improve the decision-making efficiency.

[0058] Furthermore, this embodiment provides a step for obtaining the excretion amount and remaining bait amount of each type of fish species in each fish pond within a fixed time, including:

[0059] Obtaining the feeding amount, feeding rate, absorption rate of each type of fish species in each fish pond within a fixed time, and the cleaning times and water change times of each fish pond;

[0060] Obtaining the remaining bait amount according to the feeding amount, feeding rate, cleaning times and water change times;

[0061] Obtaining the excretion amount according to the feeding amount, absorption rate, cleaning times and water change times.

[0062] Specifically, assuming that the fixed time is 30 days, the remaining bait amount F ijx of the jth type of fish species in the ith fish pond on the xth day is:

[0063] F ijx = (F ijx-1 + f ijx (1 - r ij )) × a α × b β

[0064] Among them, F ijx-1 represents the remaining amount of bait of the j-th type of fish species in the i-th fish pond on the (x - 1)-th day, f ijx represents the amount of bait fed to the j-th type of fish species in the i-th fish pond on the x-th day, r ij represents the feeding rate of the j-th type of fish species in the i-th fish pond, α represents an indicator function related to the cleaning situation. If the i-th fish pond is cleaned on the x-th day, then α = 1; otherwise, α = 0. a represents the proportion of the uneaten bait taken away each time the cleaning is carried out to the total uneaten bait. The total uneaten bait is represented by (F ijx-1 + f ix (1 - r j ))), β represents an indicator function related to the water change situation. If the i-th fish pond is changed water on the x-th day, then β = 1; otherwise, β = 0. b represents the proportion of the uneaten bait taken away each time the water change is carried out to the total uneaten bait. Since the water change usually occurs after the cleaning, the total uneaten bait at this time should be (F ijx-1 + f ix (1 - r j )) × a α .

[0065] Among them, the feeding rate refers to the ratio of the fish's food intake to the fish's body weight. According to this definition, the specific calculation formula for the feeding rate is:

[0066] r ij = X / J

[0067] Where X represents the food intake of the j-th type of fish species in the i-th fish pond on the x-th day, J represents the body weight of the j-th type of fish species in the i-th fish pond, and the body weight of the fish species can be obtained by weighing the fish species. Specifically, the food intake is obtained based on the amount of bait fed f ijx and the remaining amount of bait F ijx-1 The remaining amount of bait can be obtained from the image. For example, the bait put in can be marked with features, such as using bait with a specific color or shape, and then taking an image of the fish pond after the fish finish eating, and finally identifying the remaining amount of bait through image recognition.

[0068] The excretion amount M of the j-th type of fish species in the i-th fish pond on the x-th day ijx is:

[0069] M ijx = (M ijx-1 + (f ijx × rij )(1 - η ij )) × c α × d β

[0070] where M ijx-1 represents the excretion amount of the j-th type of fish species in the i-th fish pond on the (x - 1)-th day, η ij represents the absorption rate of the j-th type of fish species in the i-th fish pond, c represents the proportion of the excrement taken away each time during cleaning in the total excretion amount, and the total excretion amount is represented by (M ijx-1 + (f ijx × r ijk )(1 - η ij )); d represents the proportion of the excrement taken away each time during water change in the total excretion amount, and the total excretion amount at this time is (M ijx-1 + (f ijx × r ij )(1 - η ij )) × c α .

[0071] The absorption rate reflects the proportion of the part that can be effectively absorbed and utilized by the fish species from the ingested bait in the food intake. For example, if the absorption rate of a certain fish is 0.8, it means that it can absorb 80% of the components in the ingested bait, and the remaining 20% is excreted in the form of excrement, etc. The specific formula is:

[0072] η ij = I / X

[0073] where I represents the total amount of the part absorbed by the fish species from the bait. Specifically, I can be obtained from the ratio of the fish species weight gain and the nutritional conversion coefficient. The fish species weight gain can be obtained by weighing the fish species, and the nutritional conversion coefficient represents the mass of the fish species weight converted from each unit mass of bait. For example, if the nutritional conversion coefficient is 0.4, it means that for every 1 gram of bait absorbed by the fish species, the fish species weight gain is 0.4 grams.

[0074] The nutritional conversion coefficient can be obtained from simulation experiments. Specifically, based on the daily food intake and the protein content in the bait, the total amount of protein ingested by the fish species during the experiment can be obtained. Then, based on the amount of feces collected during the experiment and the protein content in the feces, the total amount of protein excreted by the fish through feces during the experiment can be calculated. Based on the total amount of ingested protein and the total amount of excreted protein, the total amount of absorbed protein can be obtained. Then, calculate the weight gain of the fish body during the experiment. The ratio of the fish body weight gain to the total amount of absorbed protein is the protein conversion coefficient; among them, the protein content in the feces can be measured using methods such as the Coomassie Brilliant Blue method and the Kjeldahl method.

[0075] Therefore, for different fish species at different growth stages, simulation experiments can be carried out separately to obtain the absorption rates corresponding to different fish species. Moreover, only one feasible implementation is given in this embodiment, but this embodiment is not limited thereto. Those skilled in the art can make further improvements based on this embodiment to obtain a more accurate absorption rate. For example, only one nutrient, protein, is considered in this embodiment, but in actual applications, other nutrients such as fat and carbohydrates can be further considered.

[0076] It should be noted that the technical contribution of this application does not lie in the accurate determination of the absorption rate and feeding rate. The absorption rate and feeding rate of this application can be within the error range allowed by the industry. At the same time, the determination of the above absorption rate and feeding rate belongs to the well-known technology in this field, and this application does not make any restrictions or elaborations on it.

[0077] Moreover, a, b, c, and d need to be determined according to the actual breeding situation. For example, if the water flow is gentle and there is a filtration measure during water change, the values of b and d should be in the range of 0.1 - 0.3. If the water change flow rate is large and the water flow is relatively rapid, the value of b should be in the range of 0.4 - 0.6, and the value of d should be in the range of 0.3 - 0.5. If professional equipment such as a siphon is used for cleaning, the value of a should be in the range of 0.3 - 0.6, and the value of c should be in the range of 0.6 - 0.8. If only simple manual fishing and cleaning are carried out, the value of a should be in the range of 0.1 - 0.3, and the value of c should be in the range of 0.3 - 0.5.

[0078] Based on the feeding rate and absorption rate of fish species, as well as the water change and cleaning conditions of the fish pond, this embodiment establishes a formula for the remaining bait amount and excretion amount. At the same time, this embodiment takes into account that in some large-scale aquaculture fish ponds, the water flow speed is relatively slow during water change, making it difficult to wash up and carry away the bait or excrement deposited in the low-lying areas at the bottom of the pond or covered areas, and some bait has certain viscosity or water absorption, making it easy to adhere to the bottom of the fish pond or other object surfaces and difficult to be washed away by the water flow. Also, due to the limitations of cleaning tools and methods, common fishing net tools cannot completely clean up fine bait and excrement, and siphon devices are difficult to clean the bait firmly adsorbed on the bottom or wall of the pond, and the bait or excrement located at the suction port and hard-to-reach corners. Therefore, by setting an indicator function and the proportion of bait and excrement taken away by each cleaning and water change, this embodiment can more accurately calculate the remaining bait amount and excretion amount.

[0079] Furthermore, this embodiment provides a step for predicting the ammonia nitrogen content in each fish pond within a preset time according to the excretion amount, remaining bait amount, and long-short term sequence model, including:

[0080] Training a long-short term sequence model corresponding to each type of fish species according to historical breeding data;

[0081] The excretion volume and remaining bait volume of each type of fish in each fish pond within a fixed time are input into the long-short time series model corresponding to each type of fish, and the ammonia nitrogen content in each fish pond within a preset time is predicted.

[0082] In order to improve the accuracy of the prediction, the historical breeding data used for training should be breeding data over a longer period of time. For example, the historical breeding data include the daily feed residue, excretion and ammonia nitrogen content of each type of fish species within one year.

[0083] This embodiment uses a long-short time series model for prediction, which can more accurately capture the time series characteristics of historical breeding data and make accurate predictions based on the time series characteristics.

[0084] Furthermore, this embodiment provides a step of training a long-short time series model corresponding to each type of fish species based on historical breeding data, including:

[0085] Divide historical breeding data into validation set and test set;

[0086] According to the validation set, the initial long-short time series model is trained to obtain the optimized long-short time series model;

[0087] The optimized long-short time series model is evaluated according to the test set to obtain the long-short time series model corresponding to each type of fish species.

[0088] Among them, in order to ensure that each set has enough data for the corresponding task, this embodiment sets the ratio of the data in the validation set and the test set to the total data to 7:3. Without making the validation set too scarce, sufficient data can be reserved for the test set to optimize the performance of the long-short time series evaluation model, avoiding the situation where the evaluation results are inaccurate or unrepresentative due to too little data.

[0089] In addition, before training, this embodiment needs to perform initial design of the model in advance to obtain an initial long-short time series model. The long-short time series model includes an input layer, an LSTM layer, and at least one fully connected layer. The number of neurons in the input layer is equal to the number of input features. The features selected in this embodiment are the excretion volume and the remaining amount of bait. The number of neurons in the LSTM layer and the fully connected layer can be adjusted according to actual conditions, but the number of neurons in the last fully connected layer is determined according to the preset time. For example, when the preset time is 5 days, that is, when the ammonia nitrogen content of the next 5 days needs to be predicted, the number of neurons in the last fully connected layer is 5. The initial values ​​of the weight parameters and bias parameters in the model can be set by methods such as Xavier initialization, He initialization, and random initialization.

[0090] Furthermore, this embodiment provides a step of training the initial long-short time series model according to the validation set to obtain an optimized long-short time series model, including:

[0091] Based on historical aquaculture data, obtain the relationship curve between the excretion amount, the remaining bait amount, and the ammonia nitrogen content;

[0092] Based on the relationship curve, obtain the prediction training set;

[0093] Based on the prediction training set, train the initial long short-term memory model to obtain a trained model;

[0094] Based on the trained model and the validation set, obtain the optimized long short-term memory model.

[0095] Among them, the ammonia nitrogen content is the dependent variable, and the excretion amount and the remaining bait amount are the independent variables. Based on historical aquaculture data, the method for obtaining the relationship curve between the excretion amount, the remaining bait amount, and the ammonia nitrogen content can be linear fitting, polynomial fitting, etc., and this embodiment does not limit this. Then, select multiple sample points in the relationship curve. The number of sample points is the same as the number of samples in the historical aquaculture data, and the interval between each sample point is the same. Take the remaining bait amount and the excretion amount corresponding to the multiple sample points as the prediction training set.

[0096] Furthermore, this embodiment provides a step of training the initial long short-term memory model based on the prediction training set to obtain a trained model, including:

[0097] Divide the prediction training set into multiple training groups, and arrange the multiple training groups in sequence to obtain a second queuing sequence;

[0098] Perform a second iterative operation. The second iterative operation includes inputting the first training group into the current long short-term memory model for forward propagation calculation, calculating the loss function according to the forward propagation calculation result and the historical aquaculture data, calculating the gradient of the parameters in the current long short-term memory model according to the loss function and the backpropagation algorithm, updating the current long short-term memory model according to the gradient to obtain an updated long short-term memory model, updating the second queuing sequence until the second termination condition is met, and outputting the current updated long short-term memory model as the trained model; the first training group is the training group at the head of the second queuing sequence; the loss function is a regularization loss function.

[0099] Among them, the second termination condition is that the second queuing sequence is an empty sequence. The step of updating the second queuing sequence is to remove the current first training group to obtain an updated queuing sequence. And, in the second iterative operation, when performing the first iteration, the current long short-term memory model is the initial long short-term memory model, and in each subsequent iteration, the current long short-term memory model is the updated long short-term memory model obtained through the previous iteration.

[0100] Specifically, first, input the first training set into the current long short-term memory (LSTM) model. The data in the first training set are sequentially propagated forward through the LSTM layer and the fully connected layer. The forward propagation calculation result is output through the fully connected layer, that is, the ammonia nitrogen content predicted based on the remaining bait amount and excretion amount corresponding to multiple sample points. Then, calculate the loss function based on the predicted ammonia nitrogen content and the ammonia nitrogen content corresponding to multiple sample points in the relationship curve. Starting from the loss function, calculate the gradient of the loss function with respect to the parameters (weights and biases) of the last layer (fully connected layer) of the model according to the chain rule, and backpropagate the gradient to the LSTM layer to obtain the gradient of the loss function with respect to all parameters of the model (including the weights and biases of the LSTM layer and the fully connected layer). Then, update the model parameters in combination with the gradient descent method to obtain the updated long short-term memory model.

[0101] Among them, the loss function can use the regularization loss function L, and the specific formula is:

[0102]

[0103] n represents the number of predicted ammonia nitrogen contents. For example, when predicting the ammonia nitrogen contents for the next 5 days, the number of predicted ammonia nitrogen contents is 5, that is, n = 5, y q represents the q-th predicted ammonia nitrogen content, represents y q the corresponding ammonia nitrogen content in the relationship curve, λ represents the regularization parameter, ∑ p θ p is the regularization term, which means summing the squares of all parameters of the model. Its function is to keep the parameters of the model as small as possible, avoid the model from being too complex, and thus reduce the risk of overfitting. θ p represents any parameter in the current model, and p represents an index used to traverse all parameters of the model.

[0104] In this embodiment, the regularization loss function is selected for calculation. By adding a regularization term to the basic loss function, the complexity of the model can be effectively constrained to prevent the model from overfitting and improve the generalization ability of the model.

[0105] Furthermore, this embodiment provides a step of obtaining an optimized long short-term memory model based on the training model and the validation set, including:

[0106] Divide the parameters in the training model into multiple parameter groups, and arrange the multiple parameter groups in sequence to obtain the first queuing sequence;

[0107] Perform the first iterative operation. The first iterative operation includes solidifying all parameter groups except the first parameter group, updating the parameters in the current training model according to the validation set until the training model meets the first termination condition, updating the first queuing sequence, and repeating the above solidification and update steps until the first queuing sequence becomes an empty sequence, and outputting the current training model as the optimized long short-term sequence model; the first parameter group is the parameter group at the head of the first queuing sequence.

[0108] Among them, the specific steps for updating the parameters in the training model according to the validation set are as follows: divide the validation set into multiple validation groups, input each validation group into the current training model in turn for forward propagation and backpropagation calculations, and update the parameters of the current training model based on the backpropagation results, and output the updated training model. The first termination condition is that all validation groups have completed the above input and output steps once. For the first validation group to perform the input step, the current training model is the training model obtained by training the initial long short-term sequence model according to the prediction training set. For subsequent validation groups, the current training model is the updated training model obtained from the adjacent previous validation group. Moreover, in the first iterative operation, for each iterative operation, the adjacent next iterative operation is performed based on the updated training model obtained from the last validation group in the adjacent previous iterative operation, that is, the current training model in the adjacent next iterative operation is the updated training model obtained from the last validation group in the adjacent previous iterative operation. The step of updating the first queuing sequence is to remove the current first parameter group to obtain the updated first queuing sequence.

[0109] Finally, evaluate the optimized long short-term sequence model through the test set to obtain the long short-term sequence model corresponding to each type of fish species. Specifically, first input the test set into the optimized long short-term sequence model to obtain the predicted ammonia nitrogen content, and calculate the evaluation index through the predicted ammonia nitrogen content. The evaluation index can be selected as the mean square error or the mean absolute error, etc. If the root mean square error or the mean absolute error on the test set is small, it indicates that the model can better adapt to time series data and can be put into use, and then obtain the long short-term sequence model corresponding to each type of fish species; if the root mean square error or the mean absolute error is small, it indicates that the model fitting effect is poor and the model design needs to be readjusted and retrained.

[0110] In this embodiment, multiple smooth sample points are obtained through the relationship curve. Model training based on multiple smooth points can enable the long-short term sequence model to converge relatively quickly and obtain the training model rapidly. However, there are errors between the sample points obtained by fitting the relationship curve and the real data, resulting in a relatively low accuracy of the training model obtained at this time. In this embodiment, the validation set is further used to adjust the parameters in the training model and evaluate through the test set, so as to accurately obtain the long-short term sequence model corresponding to each type of fish species. Compared with directly training based on historical aquaculture data, the method provided in this embodiment uses relatively smooth sample points, avoiding the problems of poor data quality and slow convergence speed caused by the existence of outliers in historical aquaculture data, and avoiding the problem of low model accuracy through parameter adjustment and model evaluation using the validation set and the test set.

[0111] Further, this embodiment provides steps for determining the bacteria-injecting pond, the amount of bacteria injection, and the types of bacteria to be injected in each fish pond according to the ammonia nitrogen content, including:

[0112] Determine the bacteria-injecting pond according to the ammonia nitrogen content and the preset threshold;

[0113] Determine the types of bacteria to be injected according to the types and growth stages of fish species in the bacteria-injecting pond. The types of bacteria to be injected include nitrifying bacteria, denitrifying bacteria, and anaerobic ammonium oxidation bacteria;

[0114] Calculate the amount of bacteria injection according to the denitrification ability, action time, and safety factor corresponding to the types of bacteria to be injected.

[0115] Specifically, according to the prediction results of the long-short term sequence model, the ammonia nitrogen content in each fish pond within the preset time can be obtained. Compare the ammonia nitrogen content with the preset threshold. If the ammonia nitrogen content in the fish pond will exceed the preset threshold within a short time, then determine this fish pond as the bacteria-injecting pond that needs to inject denitrifying bacteria, where the value of the preset threshold and the length of the short time are set according to the actual situation.

[0116] Then, determine the types of bacteria to be injected according to the types and growth stages of fish species in the bacteria-injecting pond. The types of bacteria to be injected include nitrifying bacteria, denitrifying bacteria, and anaerobic ammonium oxidation bacteria, etc. For example, nitrifying bacteria are suitable for the fry, juvenile, and adult stages of Cyprinidae fish (such as carp, crucian carp, etc.), Perciformes fish (such as bass, grouper, etc.), and Salmonidae fish (such as salmon, etc.); denitrifying bacteria are suitable for the juvenile, adult, and breeding stages of ornamental fish with relatively high water quality requirements (such as goldfish, tropical ornamental fish, etc.) and some high-grade food fish (such as Sparidae fish, etc.); anaerobic ammonium oxidation bacteria are suitable for some fish species that are tolerant of low oxygen environments, such as loach, rice field eel, etc.

[0117] Finally, calculate the amount of bacteria injection for each fish pond according to the denitrification ability, action time, and safety factor corresponding to the types of bacteria to be injected. Exemplarily, assume that the i-th fish pond needs to inject nitrifying bacteria, and the volume of the pond water in the fish pond is Vi , the predicted ammonia nitrogen content is T i , the target ammonia nitrogen content is Then the ammonia nitrogen removal amount t i is:

[0118]

[0119] wherein, V i is in liters, T i and are in mg / L, and t i is in grams.

[0120] According to the ammonia nitrogen removal amount t i , the amount of bacteria inoculated in the i-th fish pond can be obtained as

[0121]

[0122] wherein, u represents the denitrification ability of nitrifying bacteria, with the unit of mg / L·day, u d represents the action time of nitrifying bacteria, with the unit of day, and o represents the safety factor of nitrifying bacteria, which is used to reduce the risk of unqualified denitrification or fluctuating treatment effect caused by various uncertain factors during the denitrification process. Preferably, the value range of the safety factor is 1.2 - 1.5.

[0123] Furthermore, the embodiments of the present application provide a step of quantitatively inoculating bacteria into the inoculation pond according to the amount of bacteria inoculated and the type of bacteria inoculated, including:

[0124] Determine the comprehensive score of the inoculation pond according to the amount of bacteria inoculated, the type of bacteria inoculated, the type of fish species, and the location of the inoculation pond;

[0125] Complete the quantitative inoculation of the inoculation pond in sequence according to the comprehensive score.

[0126] Specifically, the scoring criteria can be determined according to the amount of bacteria inoculated, the type of bacteria inoculated, the type of fish species, and the location of the inoculation pond. Exemplarily, for the type of fish species, sensitive fish species get 3 points, and general fish species get 1 point; for the type of bacteria inoculated, inoculating a highly targeted bacterial agent gets 3 points, and a common bacterial agent gets 1 point; for the amount of bacteria inoculated, a large demand for the amount of bacteria inoculated gets 3 points, a moderate amount gets 2 points, and a small amount gets 1 point; for the location of the inoculation pond, a location close to the bacteria tank gets 3 points, and a general location gets 1 point. According to the above scoring criteria, the comprehensive score of each inoculation pond can be determined, and the comprehensive scores are sorted in descending order to obtain the inoculation sequence of each inoculation pond.

[0127] In this embodiment, the type and growth stage of fish species, as well as the denitrification ability, action time, and safety factor corresponding to each type of bacteria, are used to accurately calculate the amount of bacteria to be added to each bacteria injection pool, and a scoring standard is established. The injection order of bacteria is determined through the comprehensive score of each bacteria injection pool, thus efficiently completing the bacteria injection work.

[0128] Furthermore, an embodiment of the present application provides a denitrifying bacteria quantitative injection system for aquaculture, including:

[0129] An acquisition module for acquiring the excretion amount and remaining bait amount of each type of fish species in each fish pond within a fixed time;

[0130] A calculation module for predicting the ammonia nitrogen content in each fish pond within a preset time according to the excretion amount, remaining bait amount, and long and short time series model, and determining the bacteria injection pool, the amount and type of bacteria to be injected into the bacteria injection pool based on the ammonia nitrogen content;

[0131] A control module for completing the quantitative injection of bacteria into the bacteria injection pool according to the amount and type of bacteria to be injected.

[0132] Among them, the acquisition module and the calculation module are located in the server, and the control module is located in the control cabinet. The control cabinet controls the bacteria tank and the valves of each fish pond according to the injection order and the type and amount of bacteria required for each bacteria injection pool to complete the bacteria injection work. And, as Figures 3 - 5 shown, there is no limitation on the installation position of the bacteria tank in this embodiment.

[0133] The denitrifying bacteria quantitative injection method and system for aquaculture provided by the embodiment of the present application can predict the ammonia nitrogen content in each fish pond in real time through the long and short time series model, and reasonably plan the amount and type of bacteria to be injected in combination with the fish species type and growth stage in each fish pond. Compared with the method of determining the ammonia nitrogen content through water quality detection, it can effectively reduce the detection cost and improve the decision-making efficiency.

[0134] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0135] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 specified in one block or multiple blocks.

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 specified in one block or multiple blocks.

[0137] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications should also be regarded as within the protection scope of the present invention.

Claims

1. A method for quantitatively adding denitrifying bacteria for aquaculture, characterized in that: include: Obtain the excretion volume and bait remaining volume of each fish species in each fish pond within a fixed period of time; According to the excretion amount, the remaining amount of bait and the long-short time series model, the ammonia nitrogen content in each fish pond within a preset time is predicted; According to the ammonia nitrogen content, determining a bacteria-adding pool in each fish pond, and the amount of bacteria added and the type of bacteria added in the bacteria-adding pool; According to the amount of bacteria and the type of bacteria, quantitative bacteria are added to the bacteria pool.

2. A method for quantitatively adding denitrifying bacteria for aquaculture according to claim 1, characterized in that: The method of obtaining the excretion volume and bait remaining volume of each type of fish species in each fish pond within a fixed time period includes: Obtaining the feeding amount, feeding rate, absorption rate of each type of fish species in each fish pond within the fixed time and the number of cleaning times and water changes of each fish pond; According to the feeding amount, the feeding rate, the cleaning times and the water changing times, the remaining amount of the bait is obtained; The excretion volume is obtained according to the feeding amount, the absorption rate, the cleaning times and the water changing times.

3. A method for quantitatively adding denitrifying bacteria for aquaculture according to claim 1, characterized in that: According to the excretion volume, the remaining amount of bait and the long-short time series model, the ammonia nitrogen content in each fish pond within a preset time is predicted, including: According to the historical breeding data, the long and short time series models corresponding to each type of fish species are trained; The excretion volume and bait remaining volume of each type of fish species in each fish pond within the fixed time are input into the long-short time series model corresponding to each type of fish species to predict the ammonia nitrogen content in each fish pond within the preset time.

4. A method for quantitatively adding denitrifying bacteria for aquaculture according to claim 3, characterized in that: The long and short time series models corresponding to each type of fish species are trained based on historical breeding data, including: Dividing the historical breeding data into a validation set and a test set; According to the validation set, the initial long-short time series model is trained to obtain an optimized long-short time series model; The optimized long-short time series model is evaluated according to the test set to obtain the long-short time series model corresponding to each type of fish species.

5. A method for quantitatively adding denitrifying bacteria for aquaculture according to claim 4, characterized in that: According to the validation set, the initial long-short time series model is trained to obtain an optimized long-short time series model, including: Based on historical breeding data, the relationship curve between excretion volume, feed residue and ammonia nitrogen content was obtained; Obtain a prediction training set according to the relationship curve; According to the prediction training set, the initial long-short time series model is trained to obtain a training model; The optimized long-short time series model is obtained according to the training model and the validation set.

6. A method for quantitatively adding denitrifying bacteria for aquaculture according to claim 5, characterized in that: According to the training model and the validation set, the optimized long-short time series model is obtained, including: Dividing the parameters in the training model into a plurality of parameter groups, and arranging the plurality of parameter groups in sequence to obtain a first queuing sequence; Execute a first iterative operation, which includes solidifying all parameter groups except the first parameter group, updating the parameters in the current training model according to the verification set until the training model meets the first termination condition, updating the first queuing sequence and repeating the above solidification and updating steps until the first queuing sequence is an empty sequence, and outputting the current training model as the optimized long-short time series model; the first parameter group is the parameter group located at the first position of the first queuing sequence.

7. A method for quantitatively adding denitrifying bacteria for aquaculture according to claim 5, characterized in that: According to the prediction training set, the initial long-short time series model is trained to obtain a training model, including: Dividing the prediction training set into a plurality of training groups, and arranging the plurality of training groups in sequence to obtain a second queuing sequence; Execute a second iterative operation, the second iterative operation including: inputting the first training group into the current long and short time series model for forward propagation calculation, calculating the loss function according to the forward propagation calculation result and the historical breeding data, calculating the gradient of the parameters in the current long and short time series model according to the loss function and the backward propagation algorithm, updating the current long and short time series model according to the gradient to obtain an updated long and short time series model, updating the second queuing sequence until the second termination condition is met, and outputting the current updated long and short time series model as the training model; the first training group is the training group at the first position of the second queuing sequence; the loss function is a regularized loss function.

8. The method for quantitatively adding denitrifying bacteria for aquaculture according to claim 1, characterized in that: According to the ammonia nitrogen content, a bacteria pool is determined in each fish pond, and the amount and type of bacteria in the bacteria pool include: Determining the bacteria pool according to the ammonia nitrogen content and a preset threshold value; Determine the type of bacteria to be added according to the type and growth stage of the fish species in the bacteria-adding pond, and the bacteria-adding types include nitrifying bacteria, denitrifying bacteria and anaerobic ammonia-oxidizing bacteria; The amount of bacteria added is calculated according to the denitrification capacity, action time and safety factor corresponding to the type of bacteria added.

9. A method for quantitatively adding denitrifying bacteria for aquaculture according to claim 1, characterized in that: According to the amount of bacteria and the type of bacteria, quantitative bacteria are added to the bacteria pool, including: Determine a comprehensive score of the bacteria-dosing pool according to the amount of bacteria, the type of bacteria, the type of fish and the location of the bacteria-dosing pool; According to the comprehensive score, the quantitative dosing of bacteria in the dosing pool is completed in sequence.

10. A quantitative delivery system of denitrifying bacteria for aquaculture, characterized in that: include: An acquisition module is used to obtain the excretion volume and bait remaining volume of each type of fish in each fish pond within a fixed time; A calculation module, for predicting the ammonia nitrogen content in each fish pond within a preset time according to the excretion volume, the remaining amount of bait and the long-short time series model, and determining the bacteria pool in each fish pond, the amount of bacteria and the type of bacteria in the bacteria pool according to the ammonia nitrogen content; The control module is used to complete the quantitative feeding of bacteria into the feeding pool according to the feeding amount and the feeding type.

Images